The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze and modify a Django codebase, which is an external data source. This creates a surface for indirect prompt injection if the code contains malicious instructions in comments, docstrings, or data migrations.
Ingestion points: The agent processes models, views, and migrations from the local Django codebase (referenced throughout SKILL.md and the references/ directory).
Boundary markers: The instructions do not currently include explicit delimiters or safety instructions to prevent the agent from following prompts embedded within the code it analyzes.
Capability inventory: The agent has the ability to write files (e.g., creating migrations, updating models) and execute shell commands via the provided prek tools.
Sanitization: No sanitization or filtering of the analyzed codebase content is implemented.
[EXTERNAL_DOWNLOADS]: The rules/astgrep_scan.py script utilizes the uvx tool to download and execute the ast-grep-cli package from PyPI at runtime.
Evidence: The _astgrep_argv function in rules/astgrep_scan.py constructs a command using uvx --from ast-grep-cli==0.44.1 ast-grep.
Context: The download targets a specific, pinned version of a well-known tool from the official Python package registry, which is a standard development practice.
[COMMAND_EXECUTION]: The skill includes a Python script (rules/astgrep_scan.py) that invokes external CLI tools using the subprocess module.
Evidence: The script calls subprocess.run to execute ast-grep (or uvx) with arguments derived from the command line.
Context: This is a deterministic execution used for static analysis and linting within a development environment.

ac-django